Research map / Graph learning / Graph self-supervised learning
Generative Code Modeling with Graphs
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use graph-based attribute decoding rather than expansion-sequence decoding (ablation) | ExprGen with graphEnc context · ExprGen test (seen projects) | Acc@1: 27.1% -> 52.3% | Helped |
| Use graph-based attribute decoding rather than expansion-sequence decoding (ablation) | ExprGen with graphEnc context · ExprGen test-only (unseen projects) | Acc@1: 17.2% -> 38.8% | Helped |
| Add production and child-index labels to child edges (ablation) | graphEnc → treeDec · ExprGen test (seen projects) | Well-Typed: 49.3% -> 78.7% | Helped |
| Add production and child-index labels to child edges (ablation) | graphEnc → treeDec · ExprGen test-only (unseen projects) | Well-Typed: 41.2% -> 74.7% | Helped |
| Add expansion-sequence edges (ablation) | graphEnc → treeDec · ExprGen test (seen projects) | Acc@1: 26.8% -> 50.5% | Helped |
| Add expansion-sequence edges (ablation) | graphEnc → treeDec · ExprGen test-only (unseen projects) | Acc@1: 19.9% -> 36.0% | Helped |
| Add the full attribute-graph information beyond expansion-sequence edges (ablation) | graphEnc → syntaxDec · ExprGen test (seen projects) | Acc@1: 50.5% -> 52.3% | Helped |
| Add the full attribute-graph information beyond expansion-sequence edges (ablation) | graphEnc → syntaxDec · ExprGen test-only (unseen projects) | Acc@1: 36.0% -> 38.8% | Helped |
| Replace graph context encoding with sequence context encoding (ablation) | ourmodel decoder · ExprGen test (seen projects) | Well-Typed: 86.4% -> 53.2% | Hurt |
| Replace graph context encoding with sequence context encoding (ablation) | ourmodel decoder · ExprGen test-only (unseen projects) | Acc@1: 38.8% -> 8.4% | Hurt |
| Add attention to production selection (ablation) | ourmodel · ExprGen | production prediction | Helped |
| Add attention to variable and literal selection (ablation) | ourmodel · ExprGen | variable/literal prediction | No clear effect |
Closest papers by meaning
- Structural Language Models of Code (ICML 2020 · similarity 0.87)
- GraphCodeBERT: Pre-training Code Representations with Data Flow (ICLR 2021 · similarity 0.87)
- Learning to Represent Programs with Graphs (ICLR 2018 · similarity 0.87)
- Emergent Representations of Program Semantics in Language Models Trained on Programs (ICML 2024 · similarity 0.84)
- Diffusion On Syntax Trees For Program Synthesis (ICLR 2025 · similarity 0.84)
- LILO: Learning Interpretable Libraries by Compressing and Documenting Code (ICLR 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.